Re: [avr-chat] isp-usb

[Joerg Wunsch]

Hanns-Konrad Unger <address@hidden> wrote:

> How about a standard USB-RS-232 converter with application AVR910?
> Pros: very cheap, no smd
> Cons: ??? (what are the known problems of AVR910?)

It will be dog slow, as the AVR910 protocol is a one-byte-at-a-time
protocol.  USB however is a packet-oriented protocol, so each single
byte has to travel across the bus as a separate packet.  Also, as the
converter stands no chance of knowing whether any further data will
follow, it imposes a timeout before assembling a packet and
transmitting it, further delaying the processing.

Our computer grandfathers already knew that request-response protocols
are seriously flawed from a performance point of view, so even
protocols like UUCP or Kermit use some kind of "window" where each
side can send unacknowledged data in advance.  (TCP then improved this
method again by introducing a window that adapts itself to the latency
and packet error rate of the transmission line.)

More intelligent programming protocols like the STK500 one don't
suffer that much from being packeted into USB as they send larger
blocks of data for anything that has a larger performance impact.
(Still, reducing the device's timeout improves the transfer speed, at
the cost of cranking up the interrupt rate to be processed by the
host, even while the chip is idle.)

What's your grief with soldering SMD?  It's not all that hard, at
least not with the pitch used for the devices in question (0.8 mm).
It starts to get messy with 0.5 mm pitch, or with BGAs.

-- 
cheers, J"org               .-.-.   --... ...--   -.. .  DL8DTL

http://www.sax.de/~joerg/                        NIC: JW11-RIPE
Never trust an operating system you don't have sources for. ;-)